CENP-B Cooperates with Set1 in Bidirectional Transcriptional Silencing and Genome Organization of Retrotransposons

Abstract

Regulation of transposable elements (TEs) is critical to the integrity of the host genome. The fission yeast Schizosaccharomyces pombe homologs of mammalian CENP-B perform a host genome surveillance role by controlling Tf2 long terminal repeat (LTR) retrotransposons. However, the mechanisms by which CENP-Bs effect their functions are ill defined. Here, we show that the multifaceted roles of Abp1, the prominent member of fission yeast CENP-Bs, are mediated in part via recognition of a 10-bp AT-rich motif present in most LTRs and require the DNA-binding, transposase, and dimerization domains of Abp1 to maintain transcriptional repression and genome organization. Expression profiling analyses indicated that Abp1 recruits class I/II histone deacetylases (HDACs) to repress Tf2 retrotransposons and genes activated in response to stresses. We demonstrate that class I/II HDACs and sirtuins mediate the clustering of dispersed Tf2 retrotransposons into Tf bodies. Intriguingly, we uncovered an unexpected cooperation between Abp1 and the histone H3K4 methyltransferase Set1 in regulating sense and antisense transcriptional silencing of Tf2 retrotransposons and Tf body integrity. Moreover, Set1-mediated regulation of Tf2 expression and nuclear organization appears to be largely independent of H3K4 methylation. Our study illuminates a molecular pathway involving a transposase-containing transcription factor that cooperates with chromatin modifiers to regulate TE activities.

INTRODUCTION

Repetitive elements are prominent features of eukaryotic genomes. They are concentrated in specialized genomic landmarks, such as centromeres and telomeres, or dispersed throughout the genome, as often the case with transposable elements (TEs) and their remnants. The pervasive presence of repetitive elements is known to have a measurable impact on the expression, organization, and evolution of eukaryotic genomes (23, 30). Genes positioned within or near centromeres and telomeres are often silenced by the presence of heterochromatin containing a high concentration of repetitive elements (62). In addition, enhancer- and promoter-associated TE sequences contribute to novel regulatory controls of gene expression and organismal development that can shape the evolutionary trajectory of a lineage (19, 27). Genes of some TEs have been captured by host genomes and co-opted for new functions. For example, the catalytic core of the RAG1 recombinase, which functions in V(D)J recombination in jawed vertebrates, is thought to be derived from Transib DNA transposons (29, 66).

The fission yeast Schizosaccharomyces pombe genome contains several types of repetitive elements having organization and characteristics resembling those of higher eukaryotes. Its centromeres, telomeres, and mating-type locus contain repeats targeted for RNA interference (RNAi)-mediated heterochromatin formation (10). The genome contains 13 full-length Tf2 long terminal repeat (LTR) retrotransposons and hundreds of Tf remnants (“solo LTRs”) which preferentially target to promoters and influence regulation of RNA polymerase II (pol II) genes (6, 34, 58). In addition, fission yeast possesses three genes (abp1, cbh1, and cbh2) having homology to the mammalian CENP-B (3). Both fission yeast and mammalian CENP-Bs are believed to be derived from DNA transposases (12, 25, 59) and have been shown to bind to pericentromeric repeats and contribute to heterochromatin formation (42, 49).

Considerable progress has been made in our understanding of the molecular factors controlling the organization and assembly of chromatin (11). Similarly to higher eukaryotes, S. pombe chromatin is organized broadly into two domains, euchromatin and heterochromatin, which are distinguishable by the presence of key modified histone residues (8, 33). Euchromatin, wherein most protein-coding genes reside, is highly enriched for hyperacetylated histones catalyzed by histone acetyltransferases (HATs) and methylation of histone H3 at lysine 4 (H3K4me) mediated by Set1 (47). In contrast, heterochromatin is enriched for histones hypoacetylated through the action of histone deacetylases (HDACs) and methylation of histone H3 at lysine 9 (H3K9me) catalyzed by Clr4, homolog of human SUV39H1 (21). One of the salient features of heterochromatin in S. pombe is the involvement of RNAi in the assembly of heterochromatin (38). Accumulating evidence suggests that the RNAi machinery cooperates with DNA-binding proteins such as CENP-B, Taz1, and Atf1 to nucleate heterochromatin at centromeres, telomeres, and the silent-mating region, respectively (20). These factors recruit the histone methyltransferase Clr4 to establish H3K9me at repeats (26, 28, 42, 67). H3K9me acts as a binding anchor for HP1 proteins Swi6 and Chp2 and other chromodomain-containing proteins, including Clr4 and Chp1 (43, 48, 74). HP1 proteins, in turn, provide a platform for the recruitment of chromatin modifiers, including HDACs (18, 40, 61). Chp1, as part of an RNAi effector complex (RITS) (15, 65), facilitates amplification of RNAi activities at the targeted repeats, thereby contributing to the assembly of heterochromatin (7, 48). Intriguingly, genome-wide mapping analysis reveals little enrichment of heterochromatin and RNAi components (H3K9me, Clr4, HP1, and RITS) at Tf2 retrotransposons (10, 74), suggesting a distinct mode for the assembly of silent chromatin at these elements.

We have previously discovered a novel genome surveillance role for CENP-Bs in the regulation of TEs in fission yeast (9). S. pombe CENP-B proteins Abp1 and Cbh1 localize to all full-length Tf2 elements as well as most solo LTRs (9, 72). Loss of abp1 results in dramatic upregulation of Tf2 retrotransposons, which is further enhanced by the combinatorial loss of cbh1 or cbh2 (9). CENP-B-bound LTRs could act as versatile regulatory modules, exerting transcriptional control of nearby genes (9, 58) and affecting other DNA transactions, such as stabilizing DNA replication forks under certain genomic contexts (72). CENP-Bs also localize to non-LTR-associated promoters and regulate expression of associated genes, several of which are involved in sex-specific directed recombination (37). Most intriguingly, CENP-Bs contribute to the clustering of dispersed Tf2 elements into distinct Tf bodies (9) and thus play a role in genome organization.

The diverse roles of CENP-Bs reflect their critical importance for the survival of fission yeast: strains lacking abp1 have a slow-growth phenotype that is exacerbated by the additional loss of cbh1 and/or cbh2 (3). However, the mechanisms by which CENP-B family proteins effect their functions remain poorly understood. To begin elucidating the diverse roles of CENP-Bs, we report here our study of Abp1, whose loss results in most of the observed CENP-B mutant phenotypes. We identified two motifs located near the 3′ end of LTRs essential for Abp1 binding and repression of LTR-associated genes. Our analysis of Abp1 protein domain deletion mutants revealed that the DNA-binding, transposase, and dimerization domains of Abp1 contribute distinct functions to transcriptional regulation and genome organization. Expression profiling revealed that Abp1 and class I HDAC Clr6 and class II HDAC Clr3 repress a common set of transcriptional targets, including Tf2 retrotransposons and genes induced in response to stresses. These class I/II HDACs and class III sirtuins also contribute to the clustering of dispersed Tf2 retrotransposons. Importantly, we discovered novel roles for Set1 in the nuclear organization and silencing of Tf2 elements. Set1 localizes to and cooperates with Abp1 to repress sense and antisense transcripts of Tf2 retrotransposons. The novel repressor function of Set1 and its role in genome organization of Tf2 retrotransposons are not adversely affected in a strain lacking H3K4 methylation. Our study provides new insights into the diverse roles of CENP-Bs, and it reveals unanticipated and more complex roles for Set1 in regulating transcription, control of repetitive elements, and genome organization.

MATERIALS AND METHODS

Strain construction and experimental conditions.

Strains were generated by standard genetic methods (39). abp1 mutant strains were constructed using site-directed mutagenesis (SDM) in which genomic DNA of a wild-type strain carrying abp1-(3×)FLAG::Kanamycin was used as a template (9). A two-step SDM strategy was used to create strains carrying either deletion of Abp1 binding sites (A1 and A2) or amino substitution of lysine 4 of histone H3 for alanine (H3K4A) or arginine (H3K4R). First, the targeted element (LTR or individual histone H3 copy) is deleted and replaced with a ura4 marker to create an LTRΔ::ura4 or H3.xΔ::ura4 (where x is either 1 [hht1], 2 [hht2], or 3 [hht3]) strain. DNA fragments generated by SDM containing either the mutant Abp1 binding site (LTR-A1Δ, -A2Δ, or A1Δ A2Δ) or H3K4 mutation were transformed into the LTRΔ::ura4 or H3.xΔ::ura4 strain, respectively. Successful replacement of ura4 with the mutant LTR or H3K4 was confirmed by PCR and DNA sequencing. Standard genetic crosses were used to construct a strain in which all three copies of histone H3 contain either H3K4A or H3K4R. Western blotting and immunofluorescence (IF) analyses with anti-H3K4me1 (ab8895; Abcam), H3K4me2 (07-030; Millipore), or H3K4me3 (07-473; Millipore) antibodies were used to confirm the absence of H3K4me in H3K4A and H3K4R mutant strains. Liquid cultures were grown at 30°C in standard rich medium supplemented with 225 mg/liter adenine (YEA).

Quantitative reverse transcription real-time PCR (qRT-PCR).

RNA was isolated by a MasterPure yeast RNA purification kit (Epicentre) and converted to cDNA with Superscript III and anchored oligo(dT) primer (Invitrogen). cDNA was subjected to qPCR analysis using a DyNAzyme II PCR master mix (Finnzymes) on the Applied Biosystems 7500 fast real-time PCR system. Fold expression changes of mutant versus wild-type cells relative to act1 gene were determined using the 2^−ΔΔCT method in Microsoft Excel.

Gene expression profiling.

Batch cultures of mutant and isogenic wild-type strains were grown in parallel to midexponential phase (optical density at 595 nm [OD₅₉₅], ∼0.3 to 0.6), harvested by brief centrifugation (2,000 rpm for 2 min), decanted, and snap-frozen in liquid nitrogen. Total RNA was extracted using a hot acid phenol method (36) and purified using RNeasy columns (Qiagen). RNA (20 μg) was reverse transcribed into cDNA using anchored oligo(dT) primer and labeled and column purified using the Superscript indirect cDNA labeling system (Invitrogen). Alexa Fluor 555-coupled wild-type cDNA (200 to 300 ng) was mixed with an equal amount of Alexa Fluor 647-coupled cDNA from respective mutant cells and hybridized on a custom 4 × 44k probe Agilent tiling microarray as previously described (10). The R/Bioconductor limma package (60) was used for intra-array normalization using the loess method. Differential expression for each probe was determined by fitting a linear model following empirical Bayes standard error smoothing. Probes with an absolute log₂ fold change of ≥1.5 in the mutant versus wild-type channels from duplicate microarrays and a false discovery rate (FDR)-adjusted P value of ≤0.05 were deemed significantly different. Gene ontology (GO) enrichment analysis was performed using the R/Bioconductor GOstats package (17). All genome annotations and mappings of GO terms to features were obtained from the Sanger Centre S. pombe reference genome database (www.pombase.org).

LTR DNA-binding and Western assays.

Protein extracts were obtained from S. pombe cells (OD₅₉₅, ∼1 to 2), resuspended in HCS buffer (150 mM HEPES, pH 7.2, 250 mM NaCl, 0.1% NP-40, 20 mM each NaF and BGP, 1 mM each EDTA, dithiothreitol, and phenylmethylsulfonyl fluoride [PMSF], and a protein inhibitor tablet [Roche]), and lysed by acid-washed beads in a bead beater (three times for 30 s with a 2-min interval on ice). For the Abp1 LTR DNA binding assay, protein extracts (0.2 to 1 mg) containing Abp1-FLAG were incubated with ∼10 ng of biotinylated LTR fragments prebound to streptavidin beads for 1 h at 4°C. Beads were extensively washed with HCS buffer and subjected to polyacrylamide gel electrophoresis (NUPAGE Novex 10% BT; Invitrogen) and Western blot analyses (iBlot; Invitrogen) with anti-FLAG (M2, Sigma) antibodies. For detecting protein levels of Abp1 domain mutants, 20 μg of protein extracts from strains expressing either full-length domains or truncation of either DNA-binding, transposase, or dimerization domain of Abp1 carrying 3 copies of FLAG epitopes at the carboxy terminus was used in Western blotting with anti-FLAG antibody (sc-807; Santa Cruz Biotechnology).

ChIP.

Chromatin immunoprecipitation (ChIP) assays were performed as previously described (9) with slight modifications. Briefly, cells were treated with paraformaldehyde followed by further cross-linking with dimethyl adipimidate dihydrochloride (DMA) (10 mM in 1× phosphate-buffered saline [PBS]) at room temperature for 45 to 60 min. A 0.5% portion of the total whole-cell lysate volume was saved for the control (“input”). ChIP and input DNA were phenol-chloroform extracted using phase lock tubes (5 Prime). qPCR was performed using DyNAzyme II PCR master mix (Finnzymes) on the Applied Biosystems 7500 fast real-time PCR system. Enrichment of ChIP versus input DNA was determined using the 2^−ΔΔCT method in Microsoft Excel.

FISH.

Fluorescence in situ hybridization (FISH) assays were performed as previously described (9). Briefly, 10 ml YEA media containing 2.4 M sorbitol was added to 10-ml cultures of growing S. pombe cells (OD₅₉₅, ∼0.5 to 1). Cells were immediately cross-linked with 2.9 ml of freshly made 30% paraformaldehyde in an 18°C water bath shaker for 30 min and quenched with 1.2 ml of 2.5 M glycine. Cells were transferred to a microcentrifuge tube, subjected to cell wall digestion in 0.5 mg/ml Zymolyase solution (100T; Associates of Cape Cod) at 37°C for 30 min, blocked with PEMBAL [100 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES; pH 6.9), 1 mM EGTA, 1 mM MgSO₄, 1% bovine serum albumin (BSA), 0.1 M l-lysine] for 1 h, and then treated with RNase A (0.1 mg/ml) at 37°C for 3 h. Hybridization was carried out with 100 to 150 ng of probes in 100 μl hybridization buffer (50% formamide, 2× SSC [1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate], 5× Denhardt's solution, 10% dextran sulfate) at 40°C for 10 to 12 h. Cells were washed three times in 100 μl 2× SSC for 30 min each before mounting. For FISH probe preparation, a 3.6-kb PCR fragment corresponding to the coding region of a full-length Tf2 element was amplified by PCR using Phusion DNA polymerase (NEB) and then cloned into a Zero Blunt TOPO vector. A 2- to 3-μg portion of Tf2-orf plasmid DNA was digested in a 50-μl reaction mixture with AluI and DdeI restriction enzymes and purified with Amicon Ultra-0.5 columns (Millipore). Digested DNA was labeled by random priming (TaKaRa) with dCTP-Cy3 (GE Healthcare) and cleaned up with Amicon Ultra-4 columns (Millipore). Images were obtained by a Zeiss Axonplan 2 microscope. Volume deconvolution was performed using the Openlab 5.5 software package. The chi-square test of homogeneity was used to determine whether declustering of Tf2 elements seen in mutant cells relative to wild-type cells was significant (P < 0.05).

Microarray data accession number.

Microarray data were deposited at NCBI's GEO archive under accession number GSE39404.

RESULTS

Identification of Abp1 binding sites within LTRs.

Previous genome-wide analyses revealed Abp1 localization at Tf2 elements and solo LTRs dispersed throughout the fission yeast genome (9, 72). Abp1 enrichment at Tf2 peaks at flanking LTRs, suggesting the existence of specific sites within LTRs important for Abp1 binding. To identify such putative Abp1 binding sites, a series of biotinylated DNA segments consisting of either full-length or truncated Tf2-6 LTR sequences were used in an in vitro DNA-binding assay with streptavidin beads in the presence of Abp1 protein tagged with a 3× FLAG epitope. DNA fragments capable of pulling down Abp1 all shared a region mapping to position 249 to 302 within the LTR sequence (Fig. 1A). Analysis of binding peaks from previously reported Abp1 ChIP with microarray technology (ChIP-chip) data (9) using the DNA motif finding program Weeder (50) revealed a conserved putative motif (TAATATAATA) that falls within this region (Fig. 1B). A sequence with a perfect match to this putative motif present within the Tf2 LTR (position 273 to 282), which we termed A1, is also present within LTR sequences (235 to 244) of Tf1, a Tf LTR retrotransposon related to Tf2 (Fig. 1B). We confirmed that the A1 site is important for the in vitro binding of Abp1 to a Tf2 LTR segment (see Fig. S1A in the supplemental material). Another putative binding motif identified from ChIP-chip data, which we labeled A2 (TAATACAATA), resides slightly upstream of A1 at position 226 to 235 of Tf2 LTR (Fig. 1B).

Fig 1.

A 10-bp motif near the 3′ end of LTRs is required for Abp1 localization at a Tf2 and a solo LTR and repression of an LTR-associated gene. (A) In vitro binding assay revealed potential Abp1 binding sites near the 3′ end of the LTR. Biotinylated LTR fragments prebound to streptavidin agarose beads were incubated with protein extracts containing Abp1-FLAG proteins. Bound Abp1-FLAG proteins were detected by immunoblotting. (B) Abp1 binding sites are enriched with two 10-bp motifs. The program Weeder was used to identify putative DNA motifs from DNA sequences corresponding to the top 20 Abp1 binding peaks obtained from Abp1 chromatin immunoprecipitation (ChIP) coupled microarray (ChIP-chip) results (9). (C) Abp1 localization at Tf2-12 requires the A1 motif. ChIP analysis was performed using Abp1-FLAG epitope-tagged strains containing deletions of A1, A2, or both A1 and A2 motifs in the 5′ LTR of Tf2-12. Quantitative PCR (qPCR) was used to assay ChIP enrichment. Percent enrichment of target amplification compared to input (whole-cell extract) was calculated using the 2^−ΔΔCT method (see Materials and Methods). The black bar indicates the position of primers used for qPCR. Primers for the act1 gene promoter were included as negative controls. Error bars indicate standard deviation. n = 3. (D) Abp1 localization at a solo LTR requires the A1 motif. ChIP and qPCR were performed as described for Fig. 1C. (E) The A1 site contributes to Abp1-mediated repression of LTR-associated gene. Gene expression analysis of SPCC11E10.07c, which is located immediately downstream from a solo LTR on chromosome 3, was performed using qRT-PCR. Fold changes were normalized by act1 expression. Error bars indicate standard deviation. n = 3.

To assess the contribution of A1 and A2 to Abp1 localization at Tf2 in vivo, we performed ChIP analysis in strains lacking one or both of these sites at the two flanking LTRs of Tf2-12. Whereas removal of the A2 site had little effect, deletion of the A1 site resulted in a dramatic reduction of Abp1 binding at both the 5′ and 3′ LTRs of Tf2-12 (Fig. 1C; see Fig. S1B in the supplemental material). These results suggest that A1 acts as the primary site for the recruitment of Abp1 to Tf2 elements.

Most of the solo LTRs (63%) enriched for Abp1 binding (>2-fold) contain the A1 consensus motif. We assessed the effect of removing the A1 and A2 sites from a solo LTR located just upstream of the gene SPCC11E10.07c. Deletion of the A1 site led to a major reduction in Abp1 binding with little effect in a strain lacking an A2 site (Fig. 1D). In a strain deficient for both sites, Abp1 binding is not detectable at the solo LTR. Furthermore, reduced Abp1 binding correlates with the increased expression of the nearby gene SPCC11E10.07c, as the strain lacking A1 or both A1 and A2 shows elevated SPCC11E10.07c expression (Fig. 1E).

The three domains of Abp1 confer distinct roles to its function.

S. pombe CENP-B homologs share a similar protein architecture with that of mammalian CENP-B, reflecting their common derived origins from pogo transposases (12). These proteins possess a DNA-binding domain at the amino terminus, a transposase domain, which contains conserved residues resembling the DDE motif within the catalytic core of pogo transposases, and a dimerization domain near the carboxy terminus (Fig. 2A) (25). The roles of these domains with respect to the functions of CENP-Bs are poorly understood. We constructed mutant abp1 strains lacking one of the three domains to investigate their effect on cell growth, chromatin association, expression, and nuclear organization of Tf2 retrotransposons. Mutant Abp1 proteins lacking any of three domains are stably expressed (see Fig. S2 in the supplemental material).

Fig 2.

Common and distinct roles among Abp1 protein domains. (A) The DNA-binding and the transposase domains of Abp1 are required for normal growth and chromosome segregation. Fission yeast CENP-B proteins share a similar protein architecture with mammalian CENP-B, containing a DNA-binding domain at the amino terminus, a large transposase domain with characteristic DDE motif, and a dimerization domain at the carboxy terminus (top). Data represent the results of serial dilution analysis of wild-type (WT) and abp1 null (abp1Δ) or domain deletion mutants in nonselective (N/S) media or in the presence of thiabendazole (TBZ) (10 μg/ml) (bottom). (B) The DNA-binding domain is required for Abp1 localization at a Tf2 element. Data represent the results of a ChIP assay of strains expressing either full-length Abp1-FLAG (WT) or Abp1-FLAG lacking one of the indicated protein domains. qPCR analysis was used to quantitate Abp1 enrichment at Tf2-12 with indicated primers (black bar). Error bars indicate standard deviation. n = 3. (C) Increased Tf2 expression in abp1 mutant cells lacking the DNA-binding or transposase domain of Abp1. qPCR was performed from cDNA derived from indicated abp1 mutant strains. Error bars indicate standard deviation. n = 3. (D) All three domains of Abp1 contribute to the clustering of Tf2 elements. Data represent the results of FISH analysis of WT and indicated abp1 mutant strains using a FISH probe that corresponds to the ∼3.6-kb coding region of Tf2 elements. Representative FISH images from indicated strains are shown (top). Data represent the results of quantitative FISH analysis of observed Tf2 foci/cell in indicated strains (bar graph, bottom). n, number of cells analyzed per strain. Tf2 declustering in all abp1 mutant strains compared to WT was significant (P < 0.0005, chi-square test).

Strains lacking the abp1 gene (abp1Δ) have been shown to exhibit a slow-growth phenotype and sensitivity to the microtubule inhibitor thiabendazole (TBZ) (25). We found that strains lacking the DNA-binding domain (abp1-DBDΔ) but not the dimerization domain (abp1-DIMΔ) display growth defects and TBZ sensitivity similar to those of abp1Δ, suggesting that the lack of these defects in wild-type cells depends on Abp1 possessing an intact DNA-binding domain (Fig. 2A). Intriguingly, cells lacking the transposase (abp1-DDEΔ) domain exhibit a slight growth defect and sensitivity to TBZ (Fig. 2A). We assessed the loss of individual domains on the ability of Abp1 to associate with LTR sequences. An in vitro pulldown assay showed that whereas Abp1 proteins with the DDE truncation have slightly reduced LTR binding, the lack of either the DBD or DIM domain abrogates Abp1 binding altogether (see Fig. S3A in the supplemental material). We next determined whether the domain deletions would similarly affect Abp1 association with LTR elements in vivo using ChIP. In contrast to the in vitro results, only deletion of the DNA binding domain completely abrogated Abp1 binding at the 5′ LTR of Tf2-12, while deletion of the DDE or dimerization domain caused only a slight reduction in enrichment (Fig. 2B). Similar effects were observed at a solo LTR upstream of the SPCC11E10.07c gene on chromosome III (see Fig. S3B in the supplemental material).

We have previously shown that loss of abp1 results in dramatic increase in Tf2 expression (9). It remains unclear, however, which domain of Abp1 contributes to the repressor function of Abp1. We performed reverse transcription followed by real-time PCR (qRT-PCR) analysis to assess Tf2 expression in an abp1 mutant strain lacking one of the three domains. Consistent with our prior observation, removal of the dimerization domain has little effect on Tf2 expression (Fig. 2C) (9). However, deletion of either the DBD or DDE domain of abp1 results in a noticeable increase in Tf2 expression, but not to the same extent as a full deletion of abp1 (abp1Δ) (Fig. 2C), suggesting that DBD and DDE domains have distinct contributions to the repression of Tf2.

Abp1, in part via its dimerization domain, has a critical role in the clustering of dispersed Tf2 elements into Tf bodies (9). The effects of the other domains on Tf body formation have not been explored. We employed a fluorescent in situ hybridization (FISH) assay using a DNA probe that spans the 3.6-kb coding region of Tf2 to assess the status of Tf bodies in cells carrying abp1 domain mutants. As was noted before (9), most wild-type cells exhibit 1 to 2 Tf2 spots, while abp1Δ and abp1-DIMΔ strains display an elevated percentage of cells having two or more spots (P < 0.0005, chi-square test) (Fig. 2D). We found that strains lacking either the DBD or DDE domain also display a higher percentage of cells with two or more spots than the wild type (P < 0.0005) (Fig. 2D), suggesting that all three domains contribute to Abp1-mediated clustering of Tf2 retrotransposons.

Abp1 and HDACs target a common set of stress-response genes.

We have recently shown that Abp1 mediates Tf2 silencing in part by recruiting the HDACs Clr3 and Clr6 (9). Previous studies reported widespread derepression of genes involved in response to nutrient deprivation and other stresses in mutant strains deficient for clr6 or clr3 and clr6 (24) and a marked increase in antisense expression in clr6 mutant cells (45). We investigated the genome-wide targets of these factors by performing expression profiling in cells deficient for either abp1 (abp1Δ) or both clr3 and clr6 (clr3Δ clr6-1) using tiling microarrays. These results revealed that Abp1 acts predominantly as a transcriptional repressor and shares a common set of targets with Clr3 and Clr6 (Fig. 3A; see Table S1 in the supplemental material). Gene ontology (GO) analysis of microarray data showed significant enrichment of terms pertaining to stress response in both the abp1Δ and the clr3Δ clr6-1 arrays (see Table S2 in the supplemental material). For example, top GO terms included “cellular response to stimulus” (GO:0051716, abp1Δ P = 3.9 × 10⁻¹⁶, clr3Δ clr6-1 P = 1.2 × 10⁻⁹), “cellular response to stress” (GO:0033554, abp1Δ P = 7.1 × 10⁻¹⁶, clr3Δ clr6-1 P = 3.1 × 10⁻¹⁰), and “carbohydrate metabolic process” (GO:0005975, abp1Δ P = 7.2 × 10⁻⁶, clr3Δ clr6-1 P = 3.9 × 10⁻⁶). These same terms were significantly enriched in the set of upregulated transcripts common to both abp1Δ and the clr3Δ clr6-1 cells (P ≤ 1 × 10⁻⁶) (see Table S2), indicating that Abp1 and Clr3/Clr6 repress the same set of transcripts involved with stress response. Our results suggest that in addition to Tf2 retrotransposons and noncoding transcripts associated with heterochromatin domains, Abp1 might act as a general repressor by recruiting these HDACs to transcriptionally poised genes that function in stress response.

Fig 3.

HDACs target a shared set of genes with Abp1 and contribute to Tf body integrity. (A) Venn diagram of upregulated transcripts in abp1Δ, clr6-1, and clr3Δ clr6-1 HDAC mutants. Numbers denote significantly upregulated transcripts from duplicate microarray experiments (log₂ fold change in mutant versus wild-type [WT] strains, ≥1.5). P values were calculated using Fisher's exact test given a total of 5,337 distinct transcripts represented on the microarray. Data for clr6-1 strains were previously reported (45). (B and C) Defects in Tf body integrity in mutant cells deficient in class I (clr6) or class II (clr3) HDACs (B) or sirtuins (C). Data represent the results of FISH analyses using a Tf2 probe in WT and indicated HDAC mutant strains. Representative FISH images from indicated strains are shown (top). Data represent the results of Quantitative FISH analysis of observed Tf2 foci/cell in indicated strains (bar graphs, bottom panel). n, number of cells analyzed per strain. Tf2 declustering in mutant strains compared to that in WT strains was significant (P < 0.0005, chi-square test) except for sir2Δ (P > 0.05).

Class I/II HDACs and sirtuins contribute to Tf body integrity.

All three major classes of S. pombe HDACs play significant though disparate roles in transcriptional silencing at major heterochromatin domains and Tf2 elements (16, 20, 24). We performed FISH analysis to determine whether HDACs also contribute to Tf2 clustering. We found a noticeable increase in the number of cells with two or more spots in clr3 (class II) and clr6 (class I) mutant strains compared to the wild type (P < 0.0005, chi-square test) (Fig. 3B). Fission yeast contains class III HDAC sirtuin homologs (sir2, hst2, and hst4) that utilize NAD (NAD⁺) as cofactors (5). Previous ChIP-chip analysis shows that all three sirtuins localize to Tf2 retrotransposons and that hst4 mutant cells exhibit defects in 5′ end processing of Tf2 mRNA (16). We found that except for sir2 (P = 0.18), hst2 and hst4 mutant strains also exhibited significant defects in Tf2 clustering (P < 0.0005, chi-square test) (Fig. 3C). These results indicate that the integrity of Tf bodies requires activities from all three classes of histone deacetylases.

Novel role for Set1 in the repression of Tf2 sense and antisense transcription.

Previous whole-genome analyses have revealed little enrichment of heterochromatin markers (i.e., Swi6, Clr4, H3K9me), but noticeable levels of H3K4me, at Tf2 retrotransposons (10, 74). These findings suggest a role for the Set1 H3K4 methyltransferase, rather than the Clr4 H3K9 methyltransferase, in Tf2 regulation. To investigate this possibility, we examined the status of Tf2 expression in a set1 mutant strain using qRT-PCR. Tf2 expression was highly upregulated in cells deficient for set1 (set1Δ), to a level comparable to that of abp1Δ cells (Fig. 4A). There was a modest additional increase in Tf2 expression in cells lacking both set1 and abp1. These results suggest that abp1 and set1 have partial overlapping roles in the regulation of Tf2 elements.

Fig 4.

Set1 regulates expression and genome organization of Tf2 retrotransposons. (A) Tf2 expression is highly upregulated in a set1 mutant strain. Data represent results of qPCR analysis of Tf2 expression in strains lacking abp1, set1, or both. (B) Expression changes on sense and antisense strands of the Tf2 open reading frame (Tf2 ORF) in abp1Δ, set1Δ, and abp1Δ set1Δ mutants. Tiling microarray probes corresponding to both forward and reverse strands from each window were binned into ∼600-bp windows, and log₂ fold changes of mutant versus wild-type cells (WT) from duplicate arrays for each mutant strain in each window were averaged. (C) set1 is required for maintaining the integrity of Tf bodies. Representative FISH images from indicated strains are shown (top panel). Data represent the results of quantitative FISH analysis of observed Tf2 foci/cell in indicated strains (bar graph; bottom panel). n, number of cells analyzed per strain. Significant Tf2 declustering (P < 0.0005, chi-square test) was observed in all mutant strains.

To gain additional insights into the nature of Set1-mediated repression of Tf2 retrotransposons, we used custom tiling arrays to examine the expression of Tf2 retrotransposons on both strands in abp1Δ, set1Δ, and abp1Δ set1Δ cells. Consistent with qRT-PCR results, loss of either set1 or abp1 alone led to a marked increase in expression throughout the open reading frame of Tf2 retrotransposons on the sense strand (Fig. 4B); this upregulation was slightly increased further in double mutant abp1Δ set1Δ cells. Interestingly, we observed a modest increased expression on the antisense strand in set1Δ and abp1Δ cells. In contrast to the sense strand expression, however, there was synergistic increase in antisense expression in cells lacking both abp1 and set1 compared to that in the single mutant. These results reveal an unexpected cooperation between Abp1 and Set1 in the repression of both sense and antisense expression of Tf2 retrotransposons.

Set1 contributes to the maintenance of Tf bodies.

While factors that regulate Tf2 expression tend to also affect the integrity of Tf bodies, the two processes could be decoupled, as seen in Abp1 dimerization domain mutant cells (abp1-DIMΔ). We assessed the status of Tf bodies in cells lacking set1. The loss of set1 resulted in a higher proportion of cells exhibiting Tf2 declustering, to a degree similar to that observed in abp1Δ and abp1Δ set1Δ strains (P < 0.0005, chi-square test) (Fig. 4C). These results suggest that Abp1 cooperates with Set1 not only in transcriptional regulation of Tf2 retrotransposons but also in the nuclear organization of dispersed Tf2 elements.

H3K4me mutants do not adversely affect Tf2 silencing and Tf body integrity.

The presence of H3K4me at Tf2 retrotransposons (10) suggests that Set1 is being recruited to directly silence Tf2 retrotransposons. Indeed, ChIP analysis detects Set1 localization at the coding region of Tf2 retrotransposons (Fig. 5A). To investigate whether Set1-mediated repression of Tf2 retrotransposons requires H3K4me, qRT-PCR was performed in mutant strains in which lysine 4 of histone H3 was replaced with either alanine (H3K4A) or arginine (H3K4R). Intriguingly, whereas there was a marked increase in Tf2 expression in set1Δ cells, Tf2 expression was only slightly affected in cells containing either a H3K4A or H3K4R substitution (Fig. 5B). We next assessed the requirement for H3K4me in maintaining the integrity of Tf bodies. In contrast to the dramatic declustering of Tf2 retrotransposons observed in set1Δ cells (Fig. 4C), cells carrying either H3K4A or H3K4R substitution did not exhibit any significant degree of Tf2 declustering (P > 0.15, chi-square test) (Fig. 5C). Together, these results uncover novel functions for Set1 in the silencing and genome organization of Tf2 retrotransposons that are largely independent of the status of H3K4 methylation (Fig. 6).

Fig 5.

Set1 localizes to Tf2 retrotransposons to maintain transcriptional silencing and Tf body integrity largely independent of H3K4 methylation. ChIP analysis was performed using a strain expressing Set1-FLAG. (A) Enrichment of Set1-FLAG at Tf2 elements was assayed by qPCR using primers that correspond to the coding regions of Tf2 (Tf2-ORF). DNA corresponding to mitochondrial DNA served as a negative control. (B) Highly increased expression of Tf2 retrotransposons in strains lacking set1 (set1Δ) but not in strains containing an amino acid substitution of lysine 4 of histone H3 for alanine (H3K4A) or arginine (H3K4R). qPCR analysis of expression was performed from cDNA derived from indicated mutant strains. Error bars indicate standard deviation. n = 3. (C) FISH analyses using a Tf2 probe in wild-type (WT), H3K4A, or H3K4R mutant strains. Representative FISH images from indicated strains are shown (top panels). Data represent results of quantitative FISH analysis of observed Tf2 foci/cell in indicated strains (bar graph; bottom panels). n, number of cells analyzed per strain. Tf2 declustering in H3K4A and H3K4R mutants was insignificant (P > 0.05, chi-square test).

Fig 6.

Model for the roles of Set1 in sense and antisense transcriptional silencing of Tf2 retrotransposons and the maintenance of Tf bodies. Set1 is recruited to Tf2 retrotransposons with the help of pol II and/or transcription factors (TF) (i.e., Abp1/CENP-B or unknown TFs) to repress both sense and antisense transcription. Enrichment of H3K4me at Tf2 retrotransposons suggests heightened pol II activity during G₂ phase. Set1 probably cooperates with other factors, including Abp1 and HDACs, to silence Tf2 retrotransposons and organize dispersed Tf2 retrotransposons into Tf bodies.

DISCUSSION

Genome activities are governed primarily by the recognition of specific DNA sequences by DNA-binding proteins that play essential roles in virtually all types of DNA transactions. The fission yeast CENP-Bs participate in several DNA-directed processes, including transcription, DNA replication, and recombination (1, 9, 35, 37, 72). Extensive occupancy of Abp1 and Cbh1 at diverse genetic elements throughout the fission yeast genome supports these multifaceted roles for CENP-Bs (9, 72). However, beyond their preference for sequences often associated with autonomously replicating sequence (ARS) elements (41), the specific DNA sequences recognized by CENP-Bs are not well defined. Our results revealed a 10-bp motif (TAATATAATA), termed A1, within the 3′end of LTRs that is critical for Abp1 binding and repression of LTR-associated genes. The consensus A1 motif (Fig. 1B) is conserved in most LTRs (63%) enriched for Abp1 binding. A similar conserved motif, termed A2 (TAATACAATA), located slightly upstream of the A1 site, appears to be dispensable for Abp1 binding at Tf2-12 and the solo LTR associated with the gene SPCC11E10.07c on chromosome III. However, A2 is also highly conserved in many Abp1-associated LTRs. It is thus possible that A2 is important for Abp1 binding under other genomic contexts. The preferential binding of Abp1 for AT-rich sequences, in general, helps explain its presence throughout the AT-rich fission yeast genome (36% GC), particularly at prominent genomic landmarks with a high TA content, including LTRs (67.5%), ars (>70%) (41, 57), centromeres (67.8%), and the mating-type locus (67.7%). Considering the relatively high AT content in the genome (38% GC) of Schizosaccharomyces octosporus, which is most closely related to S. pombe (53), S. octosporus CENP-Bs might be expected to occupy as many diverse genetic elements with high AT content throughout the S. octosporus genome as those of S. pombe.

Fission yeast CENP-Bs are part of a large family of DNA-binding proteins possessing a homeodomain-like helix-turn-helix (HTH) motif involved in many cellular and developmental processes (63). In addition, fission yeast CENP-Bs contain a transposase and dimerization domain, both of which are also present in mammalian CENP-B. Our results suggest that all three domains possess distinct and yet overlapping functions that mediate the multifaceted roles of Abp1 in fission yeast. Whereas the dimerization domain (DIM) of Abp1 contributes to higher-order genome organization via clustering of dispersed Tf2 and solo LTRs (9, 64), the DNA-binding domain (DBD) facilitates recognition of its genomic targets. Surprisingly, the DIM domain also helps stabilize Abp1 association with LTR sequences. The inability of the Abp1 DIM mutant to bind to LTR sequences in vitro suggests that the dimerization domain might have a role in causing a conformational change to Abp1, thereby facilitating the recognition of LTR sequences by the DNA-binding domain. However, within an in vivo context, other factors such as nearby genetic elements and chromatin-associated factors could help stabilize the localization of the Abp1 DIM mutant at LTR-associated loci. The DBD domain likely serves as a stable protein-DNA scaffold necessary for Abp1 to carry out its functions, probably through its transposase and dimerization domains. These functions might involve short- and long-range chromatin interactions ranging from repression of local promoters to higher-order organization of Tf2 elements.

The transposase domain constitutes the largest domain of fission yeast and mammalian CENP-Bs, as well as the related mammalian jerky and tigger genes, all of which are thought to be derived from pogo transposons (12). Transposase, which has been documented as the most pervasive gene found in nature (2), encodes a DNA endonuclease critical for the transposition of DNA transposons. Whether the transposase domain of pogo-derived proteins possesses any function has not been explored. Our study reveals that the transposase domain of Abp1 has disparate contributions to the diverse functions of Abp1, including transcriptional repression and long-range chromatin interactions between distal elements. The large transposase domain might provide a recruitment platform for other factors, such as HDACs, to localize to chromatin. The function of the transposase domain might not be entirely structural but could be mediated through retention of its endonuclease activity. Transposases of pogo transposons contain a catalytic core with the characteristic residues “D, D35E” (DDE motif) which are highly conserved in pogo superfamily members (54). Interestingly, Abp1 and Cbh2 still retain two of the critical residues within the characteristic DDE motif (25), raising the possibility that fission yeast CENP-Bs still retain nuclease activity to facilitate some of their functions.

Our study identifies a novel role for set1 in multiple aspects of Tf2 regulation. Set1 is important for silencing Tf2 expression, whose levels are further elevated when set1Δ is combined with abp1Δ. Interestingly, both Abp1 and Set1 repress antisense expression of Tf2 retrotransposons, the level of which increases appreciably only in abp1Δ and dramatically in abp1Δ set1Δ. In this regard, Abp1/Set1 silencing of Tf2 sense and antisense transcription is reminiscent of the repression of forward and reverse strands of heterochromatic repeats at pericentromeres by the Clr4/RNAi pathway (10, 67). Our findings indicate that Abp1 might act in an overlapping pathway with Set1 to silence Tf2 expression on the sense strand but repress Tf2 antisense expression in parallel to an auxiliary Set1 pathway.

H3K4me is a ubiquitous euchromatic marker highly enriched at active genes (10, 51), indicative of Set1 recruitment by pol II and transcription associated-factors (32, 44). H3K4me can exist as monomethylation (H3K4me1), dimethylation (H3K4me2), or trimethylation (H3K4me3) (56). Interestingly, H3K4me2 has been shown to be enriched at Tf2 retrotransposons (10), suggesting that H3K4me is not entirely incompatible with repressed loci. H3K4me has been proposed to act as a memory of recent transcriptional events (44). Increased Tf2 expression has been observed during the G₂ phase of the cell cycle (55). Thus, the presence of H3K4me at Tf2 retrotransposons may reflect active Tf2 transcription occurring briefly within G₂. This might be similar to what has been observed with heterochromatin in which heightened RNA pol II activity during S phase provides a window of opportunity for RNAi-mediated heterochromatin assembly (14, 31).

The lack of substantial upregulation and declustering of Tf2 retrotransposons in H3K4A and H3K4R mutant cells suggests that H3K4me is not actively required per se for maintaining Tf2 silencing and Tf body integrity. However, the H3K4A or H3K4R mutation prevents both H3K4me and H3K4 acetylation (H3K4ac). H3K4me in Saccharomyces cerevisiae has been shown to limit the abundance of H3K4ac at gene promoters (22), though surprisingly, loss of set1 in S. pombe does not result in a dramatic increase of H3K4ac at the bulk histone level (70). In S. pombe, H3K4ac is mediated by the MYST family histone acetyltransferase MstI. Thus, it is possible though not exclusively true, that Set1-mediated regulation of Tf2 retrotransposons could occur via H3K4me restricting the H3K4ac activity of Mst1 at Tf2 retrotransposons.

Histone methyltransferases have been shown to target nonhistone substrates. Clr4 in S. pombe methylates both H3K9 (43) and the mRNA quality control export factor Mlo3 (75). In S. cerevisiae, in addition to H3K4, Set1 has been shown to methylate the kinetochore protein Dam1 independently of H3K4me (73). The multifaceted regulation of Tf2 retrotransposons by Set1 might occur through a combination of Set1 antagonizing MstI HAT activity and Set1 methylating a nonhistone substrate(s). A Set1-methylated substrate(s) could then directly mediate the transcriptional silencing and/or clustering of Tf2 elements. Alternatively, one or more domains of Set1 could provide a recruiting platform for Tf2 regulators.

The S. cerevisiae Set1 homolog has been shown to regulate Ty1 LTR retrotransposons (4). However, the regulation of Ty1 by budding yeast SET1 is clearly distinct from Set1-mediated regulation of Tf2 retrotransposons in S. pombe. Loss of SET1 has little effect on Ty1 expression: only in an exoribonuclease XRN1 mutant background does loss of SET1 result in increased Ty1 sense strand expression that is dependent on H3K4 methylation (4). The increase is apparently due to downregulation of a cryptic antisense transcript near the 5′LTR of Ty1 that interferes with sense strand expression (4), a phenomenon not generally observed within the fission yeast lineage (53).

Our transcriptome analysis revealed a significant number of commonly targeted transcripts in the abp1 mutant and cells deficient for either HDAC clr6 or both clr6 and clr3, including many transcripts corresponding to genes known to be involved in stress response pathways. Because Abp1 localizes to many of these loci and can physically interact with these HDACs (9), it is possible that the Abp1/HDAC complex(es) occupies promoters of genes involved in stress to keep their expression in a transcriptionally poised state and might play a role similar to that of other stress response-induced transcription factors, such as Atf1 (13, 69).

Our study implicates all three classes of HDACs, as well as Set1, in organizing dispersed Tf2 elements into Tf bodies. Strains deficient in any of the HDACs (except Sir2) or Set1 have a higher proportion of cells that exhibit Tf2 declustering than does the wild type. This result is in contrast with the effects of single HDAC mutants on Tf2 silencing (9, 16, 24): single HDAC mutants display only slight increases in Tf2 expression compared to set1Δ cells. Taken together, our data suggest that Tf2 repression is mediated primarily at the level of Set1 and that individual classes of HDACs contribute distinct and perhaps auxiliary repressive functions. HDACs preferentially target different histone residues and regulate different sets of S. pombe genes (68). It is possible that Tf2 declustering is affected by a set of acetylated histone residues targeted by three classes of HDACs that are distinct from those affecting Tf2 expression.

The clustering of Tf2 retrotransposons by CENP-B/Set1/HDACs might involve mechanisms of long-range chromatin organization similar to those associated with TFIIIC-mediated clustering of tRNA genes in S. pombe (46) and insulator and enhancer elements in higher eukaryotes (52, 71). The roles of HDACs and Set1 in mediating long-range chromatin organization are unlikely to be restricted to Tf2 clustering and may be global in scope. These global functions of Set1 and HDACs might include preventing aberrant recombination and coordination of distally located genes that are transcriptionally poised in response to genome challenges such as environmental stresses.